Connecting Claude to Your Business Tools: MCP and Connectors, Explained Simply
A plain-English look at MCP and Claude's connectors: what they actually let Claude do inside your email, calendar, files and project tools, how the permissions work, and where to start.
Five weeks into this series we've covered Cowork (Claude working on your files in the background), Claude in Chrome (Claude acting inside the browser tabs you're already logged into), how to pick a model without overthinking it, and Claude sitting inside Excel and PowerPoint. This week is about the piece that ties all of it together: how Claude actually gets access to the rest of your business's software in the first place.
The short version
- MCP (the Model Context Protocol) is the open standard Anthropic created so AI assistants can plug into outside tools. Connectors are what you actually use in Claude.
- A connector links Claude to one app, such as Gmail, Outlook, Slack or Asana. You connect it once and use it from any conversation, on any device and in Cowork.
- Google Workspace connectors can read and act (send email, book meetings, share files), asking for approval first by default.
- The Microsoft 365 connector is read-only and available on every Claude plan, including Free.
- Claude only sees what your own account can see, and you can set every tool to Always allow, Needs approval or Blocked.
- Start with one connector for one recurring task, then add more once you trust how it behaves.
What MCP and Claude connectors are (without the acronym soup)
The underlying plumbing is called MCP, the Model Context Protocol: an open standard Anthropic created so that any AI assistant can talk to any outside tool in a consistent way, rather than every company building its own one-off integration. You don't need to know what MCP stands for to use it. What you'll actually see in Claude is a connector: a specific link to one app or service (Gmail, your calendar, your project tracker) that you turn on once and then use from any conversation.
The distinction worth holding onto from the Claude in Chrome post: Chrome lets Claude act on a web page you're already logged into, inside the browser, one page at a time. A connector is a direct line from Claude into an app's actual data. It works the same way whether you're on the web, desktop, mobile or in Cowork, and it doesn't need a browser tab open at all. Different mechanism, different use case, and a business will likely end up using both.
Anthropic maintains a Connectors directory of both its own integrations and hundreds of third-party ones built by other companies. You add one from Customize > Connectors, sign in to the service, and approve what Claude is asking to access. After that, you switch it on in any conversation from the + menu.
Four Claude connectors worth knowing about
Google Workspace (Gmail, Calendar, Drive) is the obvious starting point for most small businesses. Connected, Claude can search and read your email, check and manage your calendar, and pull up documents straight out of Drive. It can also send, reply to and forward emails, create and update calendar events, and upload or share files, asking for your approval first by default. Full details are in Anthropic's Google Workspace connector guide.
Microsoft 365 (Outlook, Teams, SharePoint, OneDrive) does the equivalent reading job for a Microsoft-based business, and it's available on every Claude plan, including the free one, not just Enterprise. It's read-only: Claude can search your mail and calendar, Teams chats and meeting transcripts, and your SharePoint and OneDrive files, but it can't send email or invites or change your files. See the Microsoft 365 connector guide for the setup steps.
Slack, via Claude Tag, works a little differently: instead of asking Claude from a separate chat, you tag @Claude directly in a channel or thread. It can summarise a messy thread into what got decided and what's still open, turn chat into action items, and, with connections an admin has added, pull information from tools like your issue tracker or CRM and report back where the conversation is already happening. Claude Tag is currently in public beta; the Claude Tag guide shows the kinds of requests it handles.
Project management tools round out the set most businesses actually need day to day. Asana, Notion, Linear and Atlassian products (Jira, Confluence) are all in the directory as ready-to-connect options, letting Claude read and update tasks, tickets and docs without you switching tabs to check on them yourself. The directory guide explains how to browse what else is listed before picking one.
The benefit: each of these turns "let me go check that in the other app" into something Claude can just do as part of answering your question. That only matters once you've connected the two or three apps your business actually lives in, rather than trying to wire up everything at once.
Claude connector permissions: what it can and can't do once connected
The access question is the one that actually matters for a business owner, and the honest answer is that it's bounded by permissions you control, not a blanket key to the app. A connector signs in as you, so it can only see what your own account can already see in that service. Claude doesn't get broader access than you have.
Within a conversation, you decide whether a connector is turned on at all. For each tool, you can choose Always allow, Needs approval or Blocked, and when a tool needs approval, Claude asks before it acts, with a choice of Allow once or Always allow. You can review and tighten those settings at any time from the connector's own page under Customize > Connectors, and disconnecting it signs Claude out of the service entirely. Anthropic's guide to getting started with connectors walks through each of those settings.
Third-party connectors in the directory carry a Verified or Community label. It's worth checking before connecting anything you don't recognise, because verification is a closer review, not a full security audit.
The benefit: you're not choosing between "convenient" and "safe". The approval step means Claude can save you the manual work of checking five apps without quietly taking actions you didn't sign off on.
Where to start with Claude connectors
Don't connect everything on day one. Pick the one recurring task that currently means you (or someone on your team) manually checking an app and reporting back, like "what's on the calendar this week", "what's the status of the client project" or "anything urgent in the inbox", and connect just the one tool that task needs. Get comfortable with how the approval prompts work on that single connector before adding a second one. It's a much faster way to build trust in what Claude is actually doing than switching on the whole directory and hoping it behaves.
Bringing it together
Cowork handles background work, Chrome handles acting on logged-in web pages, and connectors handle direct access to the data in the apps your business runs on: three different doors into the same assistant, each suited to a different kind of task. Together they're what turns "Claude is a chatbot I ask questions" into "Claude actually knows what's in my calendar, my inbox and my project board".
Next week closes out this series with five prompting habits that make a real difference to the quality of what Claude gives back, whichever of these tools you're using it through. If figuring out which connector is actually worth your time is more than you want to sort out alone, that's the kind of thing I help you with, on-site across Sydney, the Central Coast and Newcastle. Book a call and we'll work it out together.
Claude connectors and MCP: common questions
What is MCP (the Model Context Protocol)?
MCP, the Model Context Protocol, is an open standard Anthropic created so that any AI assistant can connect to outside tools in a consistent way, rather than every company building its own one-off integration. Claude's connectors are built on it, so you never need to deal with MCP directly to use them.
What is a Claude connector?
A connector is a direct link between Claude and one app or service, such as Gmail, Outlook, your calendar or your project tracker. You add it once from Customize > Connectors, sign in to the service, and then turn it on in any conversation from the + menu. It works on the web, desktop and mobile apps and in Cowork.
What's the difference between a Claude connector and Claude in Chrome?
Claude in Chrome acts on a web page you're already logged into, inside your browser, one page at a time. A connector is a direct line into an app's data that works without a browser tab open, from any Claude app. Most businesses end up using both for different jobs.
Can Claude send emails and book meetings through Gmail and Google Calendar?
Yes. With the Google Workspace connectors, Claude can search, read, draft and send email, create and update calendar events, and find, upload and share Drive files. By default it asks for your approval before sending, replying, forwarding, or sharing, moving or deleting files.
Does the Microsoft 365 connector work on the free Claude plan?
Yes. The Microsoft 365 connector is available on every Claude plan, including Free. It is read-only: Claude can search your Outlook mail and calendar, Teams chats and meeting transcripts, SharePoint and OneDrive, but it can't send email or invites or change files. On Team and Enterprise plans, an Owner enables it first.
Is it safe to connect Claude to my business apps?
A connector signs in as you, so Claude only sees what your account can already see in that app. You choose per conversation whether a connector is on, and for each tool you can set Always allow, Needs approval or Blocked. Disconnecting signs Claude out of the service. Third-party connectors carry a Verified or Community label, and verification isn't a full security audit, so only connect developers you trust.
How do I add a connector to Claude?
Go to Customize > Connectors, select Discover, search for the app, select Connect to Claude, then sign in to the service and approve the access it asks for. Once it shows as Connected, turn it on in a conversation from the + menu under Connectors.
Where does Mat Tower provide on-site AI automation services?
Mat Tower works on-site directly in your office across Sydney, the Central Coast and Newcastle. For clients outside these regions, travel arrangements can be built transparently into the project scope.